1. Field of the Invention
The present invention relates to a method of manufacturing of an ultrapure colloidal silica dispersion and slurry thereof. More particularly, the present invention relates to a method of chemical mechanical polishing (CMP) the surface of a substrate using such ultrapure colloidal silica prepared according to the present invention.
2. Description of Related Art
The most common process for the preparation of colloidal silica in industry is to prepare colloidal silica particles from water glass made by fusion of natural silica sands with sodium carbonate at temperature less than 1200° C. After fusion, the fused sodium silicate is quenched and completely dissolved in water, forming water glass that is highly caustic. To process colloidal silica, the water glass is further passed through a strong acidic resin bed or column for ion exchange and converted into silicic acid. The silicic acid, normally around pH 2-3, is then placed in a container, the pH adjusted to about 8 using alkali for stabilization, and then heated to an elevated temperature, 80-100° C. for particle formation.
Depending upon the processing condition, the particle size distribution of the final product can be manipulated and controlled to be from 5 nm to about 100 nm or less. Because of the nature of the raw material, silica sands, however, the final colloidal silica from this process has more or less trace metals, such as Fe, Al, and Na, from 100 ppm to 1000 ppm or less.
Although ion exchange can remove certain amount of trace metals, impurities lower than 100 ppm is very difficult to achieve with this process even using the highest quality of natural SiO2 (see, for example, “The Chemistry of Silica,” by Ralph K. Iler, John Wiley & Sons, Inc., Ed. (1979), and U.S. Pat. No. 3,947,376).
Another approach which leads to very high purity colloidal silica is by the sol-gel process. In this process, a high purity alkoxide, such as, tetramethoxy silane (TMOS) or tetraethoxy silane (TEOS), are used as the raw material. TMOS or TEOS is dissolved in methanol or ethanol first and then mixed with deionized (DI) water for hydrolysis, with NH4OH used as the catalyst.
After the colloidal silica is formed, the solution is heated to a high temperature so that the ammonia and the organic solvent can be removed by evaporation (W. Stober, et al., J. Colloid Interface Sci., 26, 62 (1968)). The colloidal silica so processed has a very high purity because of the high purity of the raw materials.
However, this approach has several drawbacks. One is that the colloidal silica from this process is much more expensive because of the highly expensive raw materials. Secondly, large quantity of impure methanol or ethanol will be generated which is not environmental friendly. Finally, the colloidal silica can have high level of ammonia and organic solvent residual, which can be very undesirable for chemical mechanical polishing (CMP) applications.
Colloidal silica comes in different sizes and shapes. The main benefit of colloidal silica over fumed silica is that they can generate very small particles, as small as 5 to 10 nm. Also colloidal silica can be well dispersed to the primary spherical particles while fumed silica particles are always aggregated. In the area of chemical mechanical polishing (CMP), this translates to very low defectivity and high removal rates on certain metals.
Most colloidal silicas, however, are quite impure. The commercially available Ludox colloidal silica from Grace has trace metals in double digit ppm ranges. Trace metals can cause killer defectivity. This makes it unacceptable for chemical mechanical polishing (CMP) application.
Highly pure colloidal silicas can be made from TEOS or TMOS. However, these types of silica are very expensive because it requires pure raw materials and a complex manufacturing process and it generates significant amount of waste. For example, 3 parts of TEOS generates approximately 1 part of silica and two parts of impure ethanol (TEOS is composed of 28% SiO2-72% EtOH).
The fumed silicas are generally quite pure. These are solid particles ranging from 75 to 300 nm mean particle size (MPS) with primary particles size around 20 to 40 nm. But unlike colloidal silica (which are solution grown) they have to be made into chemical mechanical polishing (CMP) slurries by high shear grinding process using water, wetting and stabilizing agents. In addition these dispersions need filtration to remove large particles. Thus, although the fumed silica is low to moderate in cost, the final dispersion can be relatively expensive. Other issue with the fumed silica is they cause high CMP defectivity
In commonly assigned copending U.S. patent application Ser. No. 11/152,873, filed on Jun. 15, 2005, which is incorporated herein in its entirety, still yet another process is disclosed to produce highly pure colloidal silica dispersion. This method includes the steps of dissolving a fumed silica in an aqueous solvent containing an alkali metal hydroxide to produce an alkaline silicate solution, such as, a potassium silicate solution; removing the majority of alkali ions via ion exchange to produce a silicic acid solution; adjusting the temperature, concentration and pH of the silicic acid solution to values sufficient to initiate nucleation and particle growth; and cooling the silicic acid solution sufficiently to produce the colloidal silica dispersion. The colloidal silica particles in the colloidal silica dispersion have a primary particle size about 2 nm to about 100 nm, and a mean particle size (MPS) of 20 to 200 nm.
Semiconductor manufacturing processes, in general, use high purity chemicals, such as hydrogen peroxide, HF, TMAF, and the like. This is needed because certain impurities can affect the yield of the IC devices dramatically.
The CMP process, however, presents a different story. This is because for the 130 and 90 nm devices the amount of impurities left on the wafer surface after post-CMP cleaning is considered a critical factor. Thus a good post-CMP cleaning could remove most particles and impurities for those nodes. Thus, the impurity needs for slurries are not as stringent as other electronic chemicals. Very impure colloidal silica (e.g., Na, Al, Ca B, etc in high levels) was used for polishing of oxide layers in these nodes especially the 130 nm node.
For the 65 nm and below nodes, however, the requirements are changing dramatically. These nodes involve complex structures including Low K and ULK materials and caps. A variety of defects that were non-critical (or tolerable) for 130 nm nodes became critical killer defects. Corrosion and Fangs are 2 examples. Fangs, or seam etching at the interface between Ta and Cu at the trenches, is acceptable up to certain level for 130 nm node. For 65 nm node more than 100A fangs are unacceptable. Very low defectivity, especially scratches, also means very careful particle size and Large Particles Count (LPC) control. Impurities can have a strong influence on defectivity.
The following is a list of impurities with known or potentially damaging effects on wafers:
(1) Electrical effects: Mobile ions, such as, Na and K;
(2) Corrosion of Cu causing opens or shorts: Halides, such as, F, Cl, Br, and I, and sulfates;
(3) Fangs: solvents, alcohols, and amines; and
(4) Scratches: Al, Fe, abrasive compounds, such as, zirconia, and alumina.
In addition to these, there are indirect effects of impurities. Cross contamination is one. For example in a Cu CMP process, if step 1 has some Cl ion in the slurry, it can contaminate the step 2 slurry and change the polish performance. It should be noted that many commercial Cu and barrier slurries are relatively pure having trace metals at less than 1 ppm level.
It is an object of the present invention to solve the purity related issues by making slurries which are ultra pure, i.e., less than 200 ppb of each individual trace metal, alcohol, ammonia, certain cations, and anions, such as, Cl and F.